Thanks Helen, for a plenty of things for me to check.

Yes, this is Superserver, but I see from firebird.log that FB server
was shut down, restarted by guardian within 3 seconds and then
some users started receiving this message while others haven't
noticed the crash at all!

Did he (fbserver) do it (shutdown) on his own will or OS killed him?

Does it look like behaviour fixed in FB2? What exactly is fixed?

The firebird.log is:

SERVER (Server) Mon Mar 27 14:02:55 2006
Shutting down the Firebird service with 18 active connection
(s) to 2 database(s)

SERVER (Server) Mon Mar 27 14:02:55 2006
The database C:\PROGRAM
FILES\ISPC\PLANNER\SERVER\FDB\ATTACH.FDB was being accessed when the
server was shutdown

SERVER (Server) Mon Mar 27 14:02:55 2006
The database C:\PROGRAM
FILES\ISPC\PLANNER\SERVER\FDB\PLANNER.FDB was being accessed when
the server was shutdown

SERVER (Client) Mon Mar 27 14:02:55 2006
C:\Program Files\Firebird\Firebird_1_5\bin\fbserver.exe:
normal shutdown


SERVER (Client) Mon Mar 27 14:02:55 2006
Control services error 1061

SERVER (Server) Mon Mar 27 14:02:55 2006
Database: C:\PROGRAM
FILES\ISPC\PLANNER\SERVER\FDB\PLANNER.FDB
internal gds software consistency check (Too many savepoints
(287))

SERVER (Client) Mon Mar 27 14:02:56 2006
INET/inet_error: send errno = 10054

SERVER (Client) Mon Mar 27 14:02:57 2006
INET/inet_error: send errno = 10054

SERVER (Client) Mon Mar 27 14:02:58 2006
Guardian starting: C:\Program Files\Firebird\Firebird_1_5
\bin\fbserver.exe


SERVER (Client) Mon Mar 27 14:03:02 2006
INET/inet_error: send errno = 10054

SERVER (Client) Mon Mar 27 14:03:04 2006
INET/inet_error: send errno = 10054

SERVER (Client) Mon Mar 27 14:03:05 2006
INET/inet_error: send errno = 10054

SERVER (Client) Mon Mar 27 14:03:07 2006
INET/inet_error: send errno = 10054

--- In firebird-support@yahoogroups.com, Helen Borrie <helebor@...>
wrote:

> But if this is Superserver, and you have the scenario where some

of

> your users are crashing out of their applications while the server

is

> waiting for more *network* packets to fill a large *database*

packet

> (like a blob or a large SQL statement), then everyone else who is
> waiting for something will just wait...and wait... for packets

that

> never come...until some will just time out because the client side

of

> the connection thinks the server has gone. (This problem is fixed

in

> Fb 2.0, btw.)
>
> You'll get similar behaviour if there is a NIC or router in the
> network with a faulty connection that drops out during a packet
> wait. Sometimes, too, people load up under-resourced Windows
> networks with lots of shares and big print queues that hog all the
> bandwidth and cause traffic jams.
>
> And there are more....like the DHCP service swooping around the
> network killing connections that have stayed idle too long, and
> reassigning the IP address to a different network node...that's
> another kind of forcible close that can happen.
>
> ./hb
>